Improved perturbation based hybrid firefly algorithm and long short-term memory based intelligent security model for IoT network intrusion detection

Abstract

The widespread implementation of the Internet of Things (IoT) has introduced several potential opportunities and benefits in all aspects of our life. However, regrettably, IoT is also accompanied by a range of vulnerabilities and susceptibility to attacks and anomalies. The primary goal of these attacks is to illicitly acquire confidential information from the system while also causing disruptions in system availability for authorized users. This research introduces an improved Long Short-Term Memory (LSTM) architecture designed to accurately detect attacks in an IoT environment. The hyper-parameters of LSTM are tuned employing a novel Memetic Self Adaptive Firefly Algorithm (MAFA). This research introduced a perturbation operator and integrated it into the proposed MAFA to prevent the occurrence of local optimum solutions in the standard firefly approach. With comparative assessment of the suggested methodology and other competing deep learning (DL) approaches, it has been determined that the proposed method outperforms in different performance measures including F1 score, F2 score, Fbeta score, precision, recall, ROC-AUC score and accuracy. The MAFA-LSTM methodology is superior to all other approaches studied, with an accuracy of 99.99%. It is highly efficient for accurately detecting intrusions in an IoT environment.